[raspberry-vi] Progress
- From: Mike Ray <mike@xxxxxxxxxxxxxxx>
- To: raspberry-vi@xxxxxxxxxxxxx
- Date: Wed, 09 Apr 2014 02:41:58 +0100
Hello list.
You're all very quiet.
I thought it was time to let you know how things are going here with
efforts to get stuff going. So in order of significance, from high to low:
I finally have speech-dispatcher running on both Raspbian and Arch Linux.
I have taken a long time over this for three very good reasons:
1. Using the espeak synthesizer with the 'espeakup' program to connect
speakup to espeak has proven to be unusable on the Pi.
Despite trying very hard to get latency tweaks to stop the stuttering,
still the speech is very poor with espeakup, and too much loading on the
latency, while it smooths the voice a bit, also produces a sluggish and
unresponsive keyboard.
2. And this is the most compelling reason to use speech-dispatcher; if
I am to get Orca running in LXDE, or some other X Windows configuration,
we need speech-dispatcher to be running, because Orca needs it.
3. It took ages to get it to work :(. Open Source is fantastic but
documentation is thin on the ground. Often it is a case of just
chucking stuff at a wall until something sticks.
Now a problem...pulseaudio. The default configuration of
speech-dispatcher is to use pulseaudio, which I hate with a passion.
And what's more I could not get pulseaudio to run reliably and it
constantly crashes.
ALSA? Well...I was using ALSA originally, but ALSA is not good with
more than one audio stream at a time, so listening to a screen-reader
while listening to streamed music might make the Pi choke a bit. Not
that I am encouraging you to run more than a couple of things in one go,
poor little Pi.
Then Rill and I were looking at Knoppix. This is a flavour of Linux,
not running on the Pi currently, that has some accessible variants,
notably something called Adriane. Anyway while we were playing with
Knoppix I noticed that it uses speech-dispatcher configured using an
audio library called libao. Don't ask me what the 'ao' bit is, 'audio
output' maybe?
Anyway now I have speech-dispatcher running on both Raspbian and Arch
with libao, and......drum roll....ABSOLUTELY NO stuttering!!
I have had to do some tweaking of the program 'speechd-up', a program
which connects speakup to speech-dispatcher. For example speakup was
saying 'capital' all the time for capital letters until I hacked the
source of speechd-up and stopped it.
I also made some other interesting discoveries about speech-dispatcher
that I could not find documented anywhere. Among these are:
1. Just commenting out a 'AddModule' directive in the config file does
NOT stop sd loading it. The actual module has also to be removed from
the module path, and not just renamed, otherwise sd recognises it as a
module and loads it anyway.
And also because of the stuff with triplets in path names (like
'/usr/lib/arm-linux-gnueabihf', where the 'arm-linux-gnueabihf' bit is
called a triplet), sd gets very confused aboute paths to modules and
libao plugins. I could have hacked the Makefiles to solve this but it
was easier to just make soft links to simulate the right paths.
speech-dispatcher and the espeak API. Like a lot of the programs that
call espeak, sd does not implement the full espeak API, although it does
better than most. So I have been able to make some tweaks to things
like sample-rate and buffer size in the espeak.conf file of sd.
Incidentally using sd I have been able to reverse the latency tweaks I
made to the espeak source while we were trying to use espeakup.
One problem that remains is that Emacspeak is still stuttering. This is
because Emacspeak does not use speech-dispatcher, but the espeak server
that it uses interfaces directly to espeak. I am looking into writing
an 'Emacspeak to speech-dispatcher' connector.
All this effort has meant I have been writing SD cards faster than I can
eat chocolate, and that's fast. Infact I have busted two SD cards in
the process. The little plastic combs between the gold contacts have
come away and sprung up. Rather than risk damaging the delicate card
slot on a Pi, or leaving small fragments of plastic in the card slot in
my screamingly fast desktop Linux machine, I threw these cards away. I
have now switched to using the little metal micro-SD to SD adaptors, on
the grounds that these will not break.
Because I have been configuring and reconfiguring the Pi so much, I have
written a mechanism which will run a series of scripts one after the
other. These scripts can be grouped into collections which reflect what
I am trying to achieve.
The mechanism which runs these scripts I am calling my 'sausage
machine', for reasons which should be obvious, and the individual script
packages I am calling 'sausage packs'.
I have packs for:
Arch Linux using speech-dispatcher and libao
Arch Linux using espeakup
Raspbian using speech-dispatcher and libao
Raspbian using speech-dispatcher and ALSA
Raspbian using speech-dispatcher and installing Orca on LXDE.
and some others. Some packs run after others. For example the one
which configures Arch for console speech can be followed by one which
installs LXDE and Orca etc.
Now we have had Orca speaking on Raspbian and LXDE a couple of times,
with help from Rill's young sighted friend Morgan. But I have taken a
step back from that to get speech-dispatcher running solidly on both
Raspbian and Arch.
Now we have that, with Rill's continueing help, I can go back to trying
to get Orca up and running and speaking immediately you type 'startx' to
go into X.
So far we have done this with LXDE for a couple of reasons:
1. Raspbian already has LXDE installed so it should be a matter of just
uninstalling inaccessible apps when we have Orca speaking.
2. Wikipaedia has a page comparing desktop environments and LXDE comes
out on top as the leanest. And after all, if it says that on
Wikipaedia, it must be true, huh?
3. Because both Raspbian and Knoppix use LXDE I have some reference
systems to look at.
4. The OpenBox window manager that Raspbian uses is very flexible for
use of the keyboard, e.g. not using the mouse.
So...after a lot of rambling, a summary:
I am about to unleash Arch Linux onto Rill to see if she can break it,
before I release it on the site for download. When I release it on the
site it is likely to come without Emacspeak initially, until I can get
the Emacspeak to speech-dispatcher connector written.
Then with that done, I will get back to getting Orca running
out-of-the-box on Raspbian.
I am beginning to think that given a couple of more weeks or so we can
see a few different images up on the web site for download, including
some which are talking GUIs.
And within a couple of months or so, more accessible images than you can
shake a stick at.
Mike
--
Michael A. Ray
Analyst/Programmer
Witley, Surrey, South-east UK
The box said: 'install Windows XP, 7 or better'. So I installed Linux
Interested in accessibility on the Raspberry Pi?
Visit: http://www.raspberryvi.org/
From where you can join our mailing list for visually-impaired Pi hackers
===========================================================
The raspberry-vi mailing list
Archives: //www.freelists.org/archives/raspberry-vi
Administrative contact: <mike.ray@xxxxxxxxxxxxxx>
-----------------------------------------------------------
Raspberry Pi and the Raspberry Pi logo are trademarks of the Raspberry Pi
Foundation.
This list is not affiliated to the Raspberry Pi Foundation and the views and
attitudes expressed by the subscribers to this list do not reflect those of the
Foundation.
Mike Ray, list creator, January 2013
Other related posts:
